Exam Feedback November 2018 Part 1 Exam Feedback

[flex]

Questions from MSC is about replacement and reuse of sample data - option is bootstrapping - not antithetic and control variate.

what's under MSC implied?
i have to details something: in the case of MCS or 'samplErr reducing' precense at a context, bootstrap isn't appropriate mention, b/c bootstrap is independent simulation method (differ from MC conceptually)

 

[kylian.mbappe]

The question literraly quoted the definition of bootstrapping...

 

[flex]

can somebody recall (or simply type U selected answ) accurately qstn concerned regression beta confidence interval construct (where're beta0 =arround +0.58, large sample and 2 option contain hi-bound 1.3+) can't remember all options for qstn, extra condition for beta and confidence lvl.

 

[Charvi]

I'm impressed with how much everyone remembers from the exam - I had forgotten about a lot of the questions until mentioned here. Just to add, there was also a question about the change in value of a portfolio using delta and gamma, and a straight-up calculate the standard deviation of two portfolios question. I ended up getting 0.04 for both - hopefully I didn't make a mistake.

Does anyone know if we'll only get our results 3rd January or do they sometimes come out earlier? Also, considering that we all found the exam more or less the same, do you think getting around 70 questions right now will cut it for a pass?

If I remember correctly, I got 0.04 for both as well.

 

[Vita_lee1017]

Hi, can you plz kindly explain how to calculate this question: 95% VAR for 60 days of 2 and above? I don’t know how to solve it. Really no clue
Thank you ahead~

 

[Devesh]

The number of theoretical questions were way more than expected. And yes, it was coming down to two options every now and then.
Also, the number of questions from foundation also seemed to be more than 20. Let's hope for the best!

 

[kylian.mbappe]

Hi, can you plz kindly explain how to calculate this question: 95% VAR for 60 days of 2 and above? I don’t know how to solve it. Really no clue
Thank you ahead~

You had to use a binomial distribution with parameters being n=60 and p=0,05. Compute probability of having 0 and 1 loss in 60 days, do the sum and remove that from 1.

 

[Risksage]

I was also expecting more of 'calculate' type of questions and felt less prepared on descriptive type.
On pros, I could cover more questions than expected; but on cons, I don't know if I got them right !

 

[luciedo]

Hi, can you plz kindly explain how to calculate this question: 95% VAR for 60 days of 2 and above? I don’t know how to solve it. Really no clue
Thank you ahead~

I used Poisson distribution with lamda=5. The asnwer = 1 - P(k=0) - P(k=1). However, from what I can recall, no answer exactly matched my calculation.

 

[luciedo]

You had to use a binomial distribution with parameters being n=60 and p=0,05. Compute probability of having 0 and 1 loss in 60 days, do the sum and remove that from 1.

Nice. The answer is 80.85% (C) then. I used Poisson and got 80.085%. Luckily I still marked C

 

[luciedo]

Did you all put "use of leverage" as the main difference between HF and MF ?

I did

 

[luciedo]

Hello all,
Here are the 81 questions I may remember from the exam. Please tell me if you remember extra questions

1 -Question about Implied volatility, option strikes and maturities in the B&S model
2 -"Entreprise-wide" risk management model question
3 -EL calculation with updated (increased value of loan) from 45 mln to 55 mln
4 -Swap comparative advantages with a table (I put the company with no absolute advantage as the fixed rate payer, not sure I was right)
5/6 -2 Code of Conduct questions (one with an employee using a model from former company)
7 -A t-test hypothesis with null hypothesis being as an equality (correct answer was failing to reject the null)
8 -A 95% confidence interval question linked to an hypothesis (the hypothetized value was in the interval if I remember correctly)
9 -APT question with updated factors (formula was sth like Ri = E(Ri) + aX + bY)
10 -Crack spread question with crude and heating oil gallons/barrels
11 -White noises and independant noises
12 -The operating / combined ratio question (105% being an incorrect solution)
13 -A question about credit spread risk (I put "credit spread can be easily found via the default transition matrix. Not sure..)
14 -Stack hedge question (30 vs 360 contracts).
15 -Change in price of a bond after duration and convexity effects combined
16 -Economic capital calculation where EL, UL and multiplicator were given
17 -A question about the impact of increasing level of significance for VaR and EL
18 -Number of contract to buy/sell to make a duration-hedge with a Cheapest-to-deliver bond (portfolio value, bond FV and durations were given)
19 -A question about mean-variance model components (inputs) vs other models inputs
20 -The variance of a Bernoulli distribution ( PD*(1-PD) )
21/22 -2 Bayes questions (one with a 3-states table (increase/neutral/decrease) and one with AA/BB bonds)
23 -How many additional Monte Carlo simulations required to narrow a confidence interval (think 12,000 was the correct answer)
24 -Sample standard deviation ("what is the StD of the most volatile asset ?")
25 -Cross hedge with maturity and correlation given (Zirconium)
26 -Long run volatility (not variance !) with GARCH(1,1) model
27 -EWMA updated variance (with previous vol and previous and new asset prices given)
28 -Updated Covariance with EWMA
29 -Adjusted beta hedge from 1,05 to 1,75
30 -Risk facing CCP (answer was Adverse selection)
31 -A question about whether netting through CCP is benefiting high-rated companies most
32 -A question where the correct answer was fat tail means time-varying volatility (acc. to me)
33 -A binomial one-step tree foe European call option
34 -correl + vol hedge ratio
35 -Delta normal VAR for a portfolio of long call options
36 -Gamma neutral portfolio
37/38 -R² and adj R² (and / or sqrt(R²), can't remember if this one or two different questions ; I hope it's two !)
39 -95% VAR for 60 days (using Poisson distribution)
40 -Converting 1-day 95% to 10-day 99% VAR
41 -Difference btw Mutual funds and Hedge Funds
42 -Sharpe ratio (need to compute the weighted exp return and std first)
43 -compute expected return with beta and risk free rate (CAPM question)
44 -a Multi factor question (or a second single factor question, can't remember precisely sorry for that)
45 -a forward rate calculation one year, two years from now
46 -FX forward with continuously componded rates (straigthforward exp(rDC - rFC) formula)
47 -a put-call parity question
48 -stop vs limit order question (correct answer was limit order, to me)
49 -Sortino ratio
50 -DV01 hedge ratio
51 -Loss frequency and severity distribution in Op. risks (Poisson + lognormal)
52 -Bootstrapping vs anthetic variate vs control variate (bootstrapping was correct acc. to me)
53 -Copulas question (Gaussian copulas was correct to me)
54 -WAM / WAY
55 -CPR / SMM
56 -A question about rating agencies process wrt local vs foreign debts
57 -A question about local vs foreign currency debt ratings (similar to 56 but definitely two different questions)
58 -A question about stocks price changes when rating changes
59 -Question about role of internal auditors in stress tests process (correct answer was sthg like "need to document material changes in stress test process")
60 -Second question about role of auditors wrt the board and shareholders
61 -Covariance stationary processes (answer was mean+covariance are stable over time)
62 -Values of beta in an stable AR(1) process (answer was 0,5 others choices being 0, 1 and 1,25)
63 -Probability question using binomial distribution applied to two cases and then sum them up
64 -FX loan without interests question (with useless and tricky wording)
65 -Net CCY exposure calculation and impact on gain/loss if foreign CCY is depreciating
66 -Price of a semi annual coupon bond (with price of the equivalent annual coupon bond given)
67 -question about spot rates curve vs forward rates (correct was spot curve is upward when fwd rates are higher that spot rates, not 100% sure but it sounds logical)
68 -question about which choices is relative to coherent risk measure (correct answer was p(A+B) <= p(A) + p(B) )
69 -second question about coherent risk measure in a bank but only with wording choices related to a bank's units
70 -basic risk definition (smth like "what is more likely related to a market risk ?" with answer being a drop in FX market)
71 -question about big failures and I thing LTCM was the correct answer (linked to model risk)
72 -UL calculation
73 -AIC curve question (B was correct)
74 -SPV question (a bank using SPV to transfer legal risk)
75/76 -2 quite esay questions about differences between OTC and exchanges
77 -duration hedge performing poorly when short term rates and long term rates are volatile and not correlated
78 -variation margin vs maintenance margin vs initial margin
79 -subordinated bonds being unsecured bonds with other unsecured bonds with a higher claim above them (not sure this is an acutal question)
80 -question about RiskMetrics vs Hybrid vs Historical models for VAR
81 -a question about the benefits of ES in operational risks and the fact that they give scenarios that had never happened in the past

Risk facing CCP....I chose model risk to calculate initial margin. From BT, adverse selection is CCPs
' disadvantage, not risk

 

[flex]

I used Poisson distribution with lamda=5. The asnwer = 1 - P(k=0) - P(k=1). However, from what I can recall, no answer exactly matched my calculation.

i already replied same (like) qstn prev feedback, in this topic lamda != (isn't equal) 5, however u solving approache is true.
and also i've felt regret b/c somebodies posters both can't carefull read old feedbacks and post and recall exam qstn content withOUT minimal required acuuracy.

 

[Anibal86]

You had to use a binomial distribution with parameters being n=60 and p=0,05. Compute probability of having 0 and 1 loss in 60 days, do the sum and remove that from 1.

Actually I also used this method, but after the exam, I was thinking whether to apply the Poisson with lambda = 60*0.05 = 3.
Doing this, the result would be 80.08%, which was in the possible answers.

 

[flex]

...I was thinking whether to apply the Poisson with lambda = 60*0.05 = 3.
Doing this, the result would be 80.08%, which was in the possible answers.

exactly

 

[luciedo]

i already replied same (like) qstn prev feedback, in this topic lamda != (isn't equal) 5, however u solving approache is true.
and also i've felt regret b/c somebodies posters both can't carefull read old feedbacks and post and recall exam qstn content withOUT minimal required acuuracy.

exactly

My typo mistake, lamda = 3 (60 * 5%). I got 80.085% why answer C was 80.85%

 

[kylian.mbappe]

Using Poisson, the answer is 0.80085172654, which is 80.09%...and not 80.08%. Not sure GARP are that accurate. But anyway, you cannot assume that lambda is 60*5% because that would mean that the 1-day VAR and the 60-day VAR are the same...

 

[cmfrtblynmb209]

Using Poisson, the answer is 0.80085172654, which is 80.09%...and not 80.08%. Not sure GARP are that accurate. But anyway, you cannot assume that lambda is 60*5% because that would mean that the 1-day VAR and the 60-day VAR are the same...

Actually, I think that you can assume that lambda = 60*.05. Each one-day VaR is, by definition and convention, an independent event. So every day (independent event with potential outcomes [break, no break]), there is a 5% chance of the 1-day VaR being exceeded. If you were to take a 100 day sample of 1-day VaR measurements/actuals, you would expect to see 5 breaks (100*.05). Similarly, for 60 days, you should expect to see 3 breaks of 1-day VaR at 95% (60*.05).

You're correct, of course, that you cannot assume that 1-day and 60-day VaR are the same, but I don't think that calculating lambda as discussed is making that assumption.

 

[kylian.mbappe]

Actually, I think that you can assume that lambda = 60*.05. Each one-day VaR is, by definition and convention, an independent event. So every day (independent event with potential outcomes [break, no break]), there is a 5% chance of the 1-day VaR being exceeded. If you were to take a 100 day sample of 1-day VaR measurements/actuals, you would expect to see 5 breaks (100*.05). Similarly, for 60 days, you should expect to see 3 breaks of 1-day VaR at 95% (60*.05).

You're correct, of course, that you cannot assume that 1-day and 60-day VaR are the same, but I don't think that calculating lambda as discussed is making that assumption.

Sure. However, I was more concerned about the nature of the Poisson distribution. It is generally designed to model the number of independent events given the frequency of such independent events. Binomial distribution is designed to model the number of independent events given the probability of occurence of such independent events. In that situation, I guess the Binomial distribution is slightly more adequate because the VAR gives information about the probability of occurence and not the frequency of the losses. I also think we may use Poisson, and that's why the two results are so close (80.1% vs 80.9%).

 

[cmfrtblynmb209]

I see your point. I've been thinking about this question for a while post-test, and I think you may be right - could be a case where either Poisson or Binomial will get you there (more or less) as long as you set up the problem correctly. I remember that using Poisson I didn't get an exact answer, so Binomial was probably the preferred method.